Development of Swimming Abilities in Reef Fish Larvae

MARINE ECOLOGY PROGRESS SERIES Vol. 202: 163–173, 2000 Published August 28 Mar Ecol Prog Ser

Development of swimming abilities in reef fish larvae

Rebecca Fisher*, David R. Bellwood, Suresh D. Job

Department of Marine Biology, James Cook University, Townsville, Queensland 4811, Australia

ABSTRACT: Recent studies have revealed that reef fish larvae have excellent sustained swimming capabilities and considerable potential for modifying their dispersal patterns by active swimming. However, these studies concentrate solely on the late pelagic phase. We examined the development of swimming abilities from hatching through to settlement in 3 reef fish species (Pomacentrus amboinensis, Sphaeramia nematoptera, Amphiprion melanopus). Larval rearing provided larvae at all stages of development. Experiments were conducted in flow chambers designed for measuring the critical and sustained swimming capability of young larvae. In all 3 species, critical swimming ability increased steadily with age, size, relative propulsive area and developmental stage of the larvae. In contrast, sustained swimming ability showed a marked inflection during development. Differ- ences among species throughout development appear to reflect variations in the developmental patterns of the 3 species. Propulsive area was highly correlated with swimming ability and may prove useful for estimating swimming capabilities among species. The results suggest that some species have the potential to actively modify their dispersal patterns from an early age.

KEY WORDS: Larvae · Swimming · Dispersal · Coral reef fish · Development · Morphology

Resale or republication not permitted without written consent of the publisher

INTRODUCTION 1997). Other workers, however, have suggested that larval behaviour, including vertical and horizontal Coral reef fishes are generally sedentary as adults, movement, may influence dispersal patterns (e.g. with most possessing a dispersive larval phase. This Cowen et al. 1993, Leis et al. 1996, Stobutzki & Bell- phase lasts for several days or weeks, and during this wood 1997). time they may travel considerable distances. The dis- Recent work has shown that settlement-stage larval tribution and abundance of reef fishes is significantly reef fishes have considerable swimming abilities over influenced by the dispersal of this planktonic larval both short-term and prolonged periods. Using a cur- stage (Doherty 1991). Knowledge of the extent of larval rent flume, Stobutzki & Bellwood (1994) showed that dispersal among reefs is an essential prerequisite for the short-duration swimming speeds of larvae are rel- understanding reef-community dynamics and man- atively high, Pomacentrus amboinensis, for example agement (Williams et al. 1984). There has been consid- averaging 43.7 cm s–1, or approx. 40 body lengths (BL) erable debate about the relative importance of active s–1. Leis et al. (1996) and Leis & Carson-Ewart (1997) behaviour in determining dispersal and distribution also recorded high swimming speeds, with capture- patterns of reef fish larvae. The traditional view, which release methods recording speeds of >60 cm s–1 for is still maintained by some authors, is that fish larvae reef fish larvae in the field. These speeds were main- have very little behavioural influence over their dis- tained for several minutes. The sustained swimming persal patterns and that currents are responsible for abilities of many species were also remarkable. Sto- the dispersal of larvae to and from reefs (e.g. Roberts butzki & Bellwood (1997) recorded some species swimming over 100 km in a single bout, lasting up to 10 d, at a speed of 13.5 cm s–1. These studies describe different *E-mail: [email protected] measures of swimming ability. The methods employed

by Stobutzki & Bellwood provide data on endurance rearing techniques, the aim of this study was to de- swimming which represents a maximal performance termine how swimming ability develops with age for measure of maintained swimming ability. In compari- several reef fish species. The specific questions to be son, the observations of Leis et al. were based in the addressed were: (1) How do critical and sustained field with unrestrained larvae, the larvae selecting swimming ability vary among species during larval their swimming speed over relatively short distances. development and (2) To what extent is swimming abil- Leis & Stobutzki (1997) suggested that similarities in ity related to morphological attributes of the larvae? the results of the 2 measures indicates that they provide comparable measures of overall abilities. Method- ological differences notwithstanding, both approaches MATERIALS AND METHODS have shown that larval fish are capable of considerable horizontal movement. Three species of reef fishes were examined: the The abilities of larval fishes to influence their disper- ambon damselfish Pomacentrus amboinensis (Poma- sal remains a controversial topic and an area of consid- centridae; Pomacentrinae), the coral cardinal Sphaera- erable debate (Roberts 1997, 1998, Bellwood et al. mia nematoptera (Apogonidae), and the anemonefish 1998, Sale & Cowen 1998). While many authors admit Amphiprion melanopus (Pomacentridae; Amphiprioni- that larval behaviour may be important (reviewed by nae). All were collected from the northern Great Bar- Leis 1991a), few have considered these abilities in the rier Reef. The species selected differed markedly in theories and models put forward. Several authors have terms of their larval ecology. P. amboinensis and S. attempted to explain distribution patterns without con- nematoptera have long larval durations of about 20 sidering active movement of larvae (Booth & Baretta and 24 d respectively, whereas A. melanopus has a 1994, Roberts 1997). In such cases, it has been assumed very short larval duration of approx. 9 d. Adult brood that there is no development of swimming ability stock were kept in >1000 l aquaria and were fed a diet throughout the larval period. The results of previous of chopped pilchards, prawns and Ascetes twice a day. studies have clearly demonstrated that this assumption Rearing methods follow Job & Bellwood (2000). Larvae is not valid (Leis & Carson-Ewart 1997, Stobutzki & were reared in 200 l aquaria at 27.5 to 29°C, under a Bellwood 1997). An alternative assumption that has 14:10 h light:dark photoperiod, throughout the study been made is that swimming ability develops rapidly period. The alga Nannochloropsis sp. was used to immediately prior to the late-pelagic phase and that green the water during the day. Larvae were fed this ability is used to move towards reefs at the time >52 µm-filtered, wild-caught plankton, supplemented of settlement (Wolanski et al. 1997). Such models occasionally by rotifers and Artemia sp., at prey densi- attempting to include larval behaviour have been lim- ties of between 2 and 6 ind. ml–1. ited by the lack of knowledge of swimming abilities in Swimming experiments for older larvae were carried younger larvae. The larvae used in the studies by Sto- out using the technique and swimming flume de- butzki & Bellwood (1994, 1997), Leis et al. (1996), Leis scribed by Stobutzki & Bellwood (1997). Two addi- & Carson-Ewart (1997) and Stobutzki (1998) were all tional swimming channels were designed for use with caught in light traps and represent late pelagic-stage younger larvae (Fig. 1). These channels operated at larvae (Choat et al. 1993). If swimming ability develops slower speeds (between 0 and 15 cm s–1) and had a earlier in the larval period, then the potential for hori- much wider swimming area to minimise the area influ- zontal swimming to modify dispersal patterns will be enced by boundary layers. The swimming channels of greatly underestimated in these dispersal models. the 2 chambers were 15 cm wide × 6.5 cm deep × 35 cm Before we can understand the potential importance long and 10 cm wide × 5.5 cm deep × 30 cm long, for the of active swimming in larval reef-fish dispersal, we ‘slow’ and ‘medium’ speeds, respectively. Both chan- need to determine the swimming ability of different nels could be placed an aquaria with greened water so species of reef fishes throughout the entire larval that there was no change in water quality between the period, rather than just the late-pelagic stage. Investi- experimental apparatus and the rearing tanks. Both gations into the abilities of younger stages of larval swimming channels were calibrated using video and reef fishes has been limited by difficulties in obtaining neutral-density particles. live individuals in the field in good condition. Larval Each batch of eggs from each species was raised rearing techniques, however, provide a means of from hatching through to settlement. Experiments determining the biological capabilities of larval fishes were performed throughout the larval period, with throughout development. By hatching and rearing lar- sampling days depending on the species. The first vae under controlled conditions, it is possible to obtain swimming trial was conducted on Day 1, approx. 12 h live specimens of known age and parentage, in good after hatching. Three batches of each species were condition, at all developmental stages. Using captive used for each swimming trial to allow for batch effects. Fisher et al.: Development of swimming abilities in reef fish larvae 165

Fig. 1. Swimming channel for younger larvae. (A) Lateral view of apparatus; (B) top view of apparatus in experimental tank

At each experimental age, 8 to 12 fish were used from of different species at different ages, sustained swim- each batch for the swimming trials. The mean of these ming trial speeds were set as a multiple of body length. 8 to 12 fish was used for statistical analyses. In all Fish were placed in the appropriate swimming channel experiments, larvae were transferred to the appropri- and swam at a speed equivalent to approx. 11 body ate swimming chamber and allowed to acclimatise for lengths (BL) per second until exhausted. Larvae were 2 to 15 min. Fish that exhibited symptoms of stress such observed continuously for the first 5 min, every 10 min as erratic swimming behaviour, lying on the bottom or for the first hour, and every 2 to 4 h thereafter. The total in corners, or clinging to the sides or surface of the duration recorded represented of the swimming time swimming channel were excluded from the experi- up to the last occasion on which an individual was ments. observed swimming. Actual swimming durations may The critical swimming speed gives an indication of be up to 4 h longer for the older larvae. the maximum sustainable swimming speed of a fish, Fishes that had swum, or siblings of the same age, and involves increasing the flow rate against which the were anaesthetised in chilled water and then fixed in fish is swimming in gradual increments. In the present 10% buffered formalin. After 12 to 48 h, larvae were study, larvae were subjected to incremental increases transferred to 70% alcohol and stored. Measurements in flow rates equivalent to approx. 3 BL every 2 min were taken of total length (from the tip of the caudal until they could no longer maintain position for the full fin to the tip of the upper jaw), body depth (the 2 min. The equation used to calculate the critical swim- height of the fish measured at the deepest region), ming speed (Ucrit) of larvae followed Brett (1964): body area (the entire area of the fish in lateral view excluding the fins) and total propulsive area (the area × Ucrit = U + (t/t i Ui) of the fish including the fins but excluding the head and gut region). Although larvae exhibit some degree where U = penultimate speed, Ui = velocity increment of shrinkage (approx. 10%), this varies among ages (3 BL), t = time swum in the final velocity increment, and parameters (Job & Bellwood 1996). For consis- and t i = set time interval for each velocity increment tency, all values are based on raw (non-corrected) (2 min). measurements. The Reynolds numbers (Re) for the Allometric studies have demonstrated that the high- sustained swimming trials were calculated for the est level of activity, achieved at the top speed within mean of each batch at each age using the formula: Re any one gait, is size-dependent, and thus at any given = U × L/ν (after Webb & Weihs 1986), where U = sus- absolute speed different-sized fish may not be at simi- tained swimming speed (m s–1), L = body depth (m), lar relative levels of exertion (Drucker 1996). In order and ν = kinematic viscosity of sea water (1.03 × 10–6 to compare the sustained swimming abilities of larvae m2 s–1). 166 Mar Ecol Prog Ser 202: 163–173, 2000

Scatterplots and regressions were used to investigate the relationship between swimming ability and age. Analyses were based on both age since hatching and developmental age. Developmental age provides a way of comparing larvae between species at similar ontogenetic stages and was calculated following Job & Bellwood (2000): developmental age = total age/ total larval duration, where total age = age of fish (days post-hatch) + egg duration and total larval duration = larval duration (days post-hatch) + egg duration. Scatterplots and regressions were also used to describe the development of swimming ability with increasing size. To investigate the relationship between swimming ability and changes in the shape of larvae, measurements were standardised by dividing each variable by total length. Zero-order, as well as partial correlation coefficients were used to determine which of the relative morphological variables were most useful for modelling swimming ability. For all regressions, dummy variable analysis using corner- point parameterisation was used to determine if there were significant differences in either the slopes or intercepts for each species (Silk 1976). This technique involves comparing the sum of squares obtained for the most complex model (each species has different slopes and intercepts) with simpler models (2 or 3 species have the same slope and/or intercept) using F- tests. Where multiple F-tests were used, a Bonferroni Fig. 2. Pomacentrus amboinensis (●), Sphaeramia nema- correction was applied. The assumptions of all regres- toptera (h) and Amphiprion melanopus (▲). Development sions were tested graphically, and data were log10- of critical swimming speed with age. (A) Raw values; transformed when appropriate. In the case of sustained (B) log10-transformed values swimming time, transformation was unable to linearise the relationship with all of the parameters investi- Table 1. Pomacentrus amboinensis, Sphaeramia nematoptera, gated. Consequently, statistical analyses based on the Amphiprion melanopus. Mean and maximum critical swim- assumption of linearity were not used and the develop- ming speed and sustained swimming time 12 h after hatching ment of sustained swimming ability was described and at settlement. Values in parentheses = SE (n = 3 batches using scatterplots. of 8 to 12 fish)

Species Critical Sustained –1 RESULTS speed (cm s ) swimming time (h) Mean Max. Mean Max.

Critical swimming ability increased with age for all At hatching 3 species (Fig. 2A). Initially, critical swimming ability P. amboinensis 3.5 (1.2) 6.4 0.11 (0.04) 0.5 increased slowly with age, then improved rapidly for S. nematoptera 1.8 (0.6) 4.4 0.02 (0.02) 0.2 older stages, with the exception of Sphaeramia A. melanopus 4.6 (0.2) 8.2 0.26 (0.12) 1.1 nematoptera, in which critical swimming ability in- At settlement P. amboinensis 30.3 (5.0) 48.1 90.4 (31.9) 234. creased steadily throughout development (Fig. 2A). At S. nematoptera 10.9 (2.2) 24.2 4.3 (4.3) 39.9 12 h after hatching, all 3 species had some critical A. melanopus 14.6 (4.5) 35.7 8.6 (4.8) 44.25 swimming ability, with mean critical swimming speeds ranging from 1.8 cm s–1 for S. nematoptera to 4.6 cm s–1 for Amphiprion melanopus (Table 1). The mean critical The relationship between critical swimming ability swimming speeds obtained just prior to settlement and age differed significantly among the 3 species. ranged from 10.9 cm s–1 for S. nematoptera to 30.3 cm In terms of critical swimming ability, Pomacentrus s–1 for Pomacentrus amboinensis (Table 1). amboinensis and Amphiprion melanopus have a very Fisher et al.: Development of swimming abilities in reef fish larvae 167

similar developmental rate with age (Fig. 2B), with no degree of sustained swimming ability between Days 3 significant difference between the slopes (F3, 60 = 1.15, and 5 after hatching (Fig. 3A). All 3 species had con- p = 0.34). A. melanopus, however, hatches with signif- siderably higher maximum swimming times, and the icantly greater critical swimming abilities than the mean values displayed here are clearly a conservative other 2 species (F1, 58 = 218.1, p < 0.0001) (Fig. 2B). estimate of the potential swimming abilities of the 3 Sphaeramia nematoptera, although hatching at a simi- species (Table 1). Although most sustained swimming lar state of development in terms of critical swimming values for S. nematoptera were near zero, the maxi- ability as P. amboinensis (F3, 60 = 0.15, p = 0.34), has a mum values obtained indicated that at 24 d some indi- significantly slower rate for the development of swim- viduals were able to swim for up 39.9 h (Table 1). A ming ability with age (F1, 58 = 194.5, p < 0.0001), taking similar pattern is obtained if the equivalent distance longer to reach much slower final speeds than P. that the larvae have travelled is calculated, based on amboinensis (Fig. 2B). the speeds at each age and the overall time swum All 3 species hatched with very little sustained swim- (Fig. 3B). All 3 species swam very short distances ming abilities, and each developed this ability quite immediately after hatching. A. melanopus was only differently with age. Sphaeramia nematoptera had the able to swim very short distances throughout the larval poorest sustained swimming ability throughout devel- period, although this species was able to swim around opment, exhibiting very little sustained swimming at 4 km just 5 d after hatching (Fig. 3B). In contrast, P. any age (Fig. 3A). Pomacentrus amboinensis had the amboinensis exhibited no sustained swimming abili- greatest sustained swimming ability, and this de- ties until 15 d after hatching. Although developing sus- veloped rapidly 12 to 15 d after hatching (Fig. 3A). tained swimming ability at a much slower rate, P. Amphiprion melanopus had shorter sustained swim- amboinensis was capable of swimming 12 km at 15 d ming times than P. amboinensis, with a mean of 8.6 h at post-hatch and 40 km at 20 d post-hatch (Fig. 3B). S. the oldest age compared to 90.4 h for P. amboinensis nematoptera, on average, was able to swim only very (Table 1). A. melanopus, however, did display the ear- short distances throughout the larval period (Fig. 3B). liest development of swimming ability, with some Critical swimming ability increased exponentially with developmental age in all 3 species (Fig. 4A). There was no significant differences in the slopes for all species, or in the intercepts for Sphaeramia nema-

toptera and Amphiprion melanopus (F3, 61 = 2.17, p = 0.10), indicating that all 3 species develop critical swimming abilities at a similar rate in terms of developmental age. Although Pomacentrus amboinensis was found to have a significantly different intercept

(F3, 61 = 4.9, p = 0.004), this difference was much less than that seen when the development of critical swimming ability is compared using age since hatching. The 3 species exhibited different sustained swimming abilities, even at similar developmental ages. Pomacentrus amboinensis had the greatest sustained swimming ability, followed by Amphiprion melanopus and then Sphaeramia nematoptera. The sustained swimming ability of S. nematoptera was very poor at all ontogenetic stages (Fig. 4B). Although the magni- tudes of sustained swimming ability were different for P. amboinensis and A. melanopus, these species developed sustained swimming abilities at a similar stage of ontogenetic development (Fig. 4B), in marked contrast to the onset of abilities in terms of age since hatching (Fig. 3A). Critical swimming speed increased proportionally with total length for all 3 species, but the rate of development and the values of critical swimming ability for Fig. 3. Pomacentrus amboinensis (●), Sphaeramia nematoptera (J) and Amphiprion melanopus (▲). Development of a given length differed among species. Amphiprion sustained swimming ability with age. (A) Sustained swim- melanopus increased swimming ability with increas- ming time in hours; (B) equivalent distance travelled ing length at a significantly faster rate than the other 168 Mar Ecol Prog Ser 202: 163–173, 2000

2 species (F1, 58 = 10.09, p < 0.001) (Fig. 5B). Pomacen- Partial correlation coefficients for the 3 morphologi- trus amboinensis and Sphaeramia nematoptera devel- cal variables indicated that for sustained and critical oped swimming ability at a similar rate with increasing swimming ability, relative propulsive area described length (Fig. 5B), with no detectable difference the greatest amount of unique variation in nearly all between these rates (F1, 58 = 0.37, p = 0.54). However, cases (Table 2). A plot of relative propulsive area there was a significant difference in the intercepts against critical swimming ability suggested a similar

(F2, 59 = 11.62, p < 0.0001), with P. amboinensis having relationship for all 3 species (Fig. 7A). Statistical analy- considerably better swimming abilities for a given sis indicated that there was only a marginal difference length than S. nematoptera (Fig. 5B). Sustained swim- between the slope for Pomacentrus amboinensis and ming time also increased with increasing total body the other 2 species (F4, 61 = 2.71, p = 0.04). Overall, crit- length, but only for P. amboinensis and A. melanopus. ical swimming ability increased similarly with increas- S. nematoptera exhibited very little sustained swim- ing relative propulsive area for all species. While there ming ability at any length (Fig. 6). P. amboinensis was no simple linear relationship between sustained developed sustained swimming ability at a body swimming time and relative propulsive area, there was length of around 11.5 mm, with shorter larvae exhibit- a similar pattern for all 3 species (Fig. 7B). There ing very little sustained swimming ability (Fig. 6). A. appeared to be a common value of propulsive area rel- melanopus appeared to develop sustained swimming ative to length at which sustained swimming ability abilities smoothly with increasing length, and showed first occurs for all 3 species. The Reynolds numbers some sustained swimming abilities even at a length of experienced by larvae throughout their developed only 5 or 6 mm, which is at a considerably shorter body increased steadily from hatching through to settle- length than P. amboinensis (Fig. 6). ment. For all species, a significant increase in sus-

Fig. 4. Pomacentrus amboinensis (●), Sphaeramia nema- Fig. 5. Pomacentrus amboinensis (●), Sphaeramia nematoptera (h) and Amphiprion melanopus (▲). Swimming ability toptera (h) and Amphiprion melanopus (▲). Mean critical vs developmental age. (A) Critical swimming speed; (B) sus- swimming speed vs total length. (A) Raw values; (B) log10- tained swimming time transformed values Fisher et al.: Development of swimming abilities in reef fish larvae 169

tained swimming ability was not recorded at Re values ing sustained swimming at Re values considerably below 500 (Fig. 8). However, Re values at the time lower than those for P. amboinensis (Fig. 8). when sustained swimming ability first occurs differed for each species, with Amphiprion melanopus exhibit- DISCUSSION

The 2 measures of swimming ability, critical and sustained, provide different estimates of swimming performance. Critical swimming speed provides a maximum performance measure of the speeds at which fishes can swim, as well as a means of directly comparing the development of swimming capabilities among species. In Pomacentrus amboinensis, Sphaera- mia nematoptera, and Amphiprion melanopus, critical swimming speeds increased with age, size, developmental age and relative propulsive area. The development of critical swimming ability is probably based largely on the development of muscles and other fea- tures related to locomotion, as well as the ability to use readily available energy reserves. The biological implications of critical swimming speeds relate to the ability of the larvae to cover short distances (<500 m) at

(A)

Fig. 6. Pomacentrus amboinensis (●), Sphaeramia nematoptera (h) and Amphiprion melanopus (▲). Mean sustained swimming ability vs total length. (A) Sustained swimming time in hours; (B) equivalent distance travelled

Table 2. Pomacentrus amboinensis, Sphaeramia nematoptera, Amphiprion melanopus. Partial correlation coefficients for (B) multiple regression between critical swimming speed and sustained swimming time versus body area, body depth, and propulsive area. All variables were standardised relative to total length. Values in bold indicate morphological variables with the highest partial correlation coefficients for each species. Data were log10-transformed prior to analysis

Species Body Body Propulsive area depth area

Critical swimming speed P. amboinensis 0.22 –0.07 0.44 S. nematoptera 0.67 –0.45 0.17 A. melanopus 0.09 0.13 0.73 Sustained swimming time ● P. amboinensis –0.10 –0.08 0.46 Fig. 7. Pomacentrus amboinensis ( ), Sphaeramia nema- ▲ S. nematoptera –0.05 0.08 0.21 toptera (h) and Amphiprion melanopus ( ). Swimming ability A. melanopus –0.32 –0.32 0.72 vs relative propulsive area. (A) Critical swimming speed; (B) sustained swimming time 170 Mar Ecol Prog Ser 202: 163–173, 2000

deal of variation among species, this method provides a maximum performance measure of the sustained swimming capabilities of larvae, which may reflect relative abilities of larvae to modify dispersal patterns in the field.

Among-species differences — Is it possible to predict swimming ability across different species?

The development of both critical swimming speeds and sustained swimming times of larvae showed clear differences among the 3 species. Comparable differences in the swimming ability among taxa have been Fig. 8. Pomacentrus amboinensis (●), Sphaeramia nema- reported by other authors (Leis et al. 1996, Stobutzki & toptera (h) and Amphiprion melanopus (▲). Sustained swim- Bellwood 1997, Stobutzki 1998), and have been attrib- ming time vs Reynolds number. Dotted line: Re = 200 uted to differences in developmental stage and morphology at the end of the pelagic stage (Stobutzki & high speeds (up to 48 cm s–1), over short time scales Bellwood 1997). The differences observed in the pre- (< 30 min). These abilities may be important in terms of sent study appear to reflect the larval strategies of the the potential for larvae to move between locations on a 3 species. They hatch at varying stages of development small scale (e.g. within slicks), to vertically migrate to and go through the larval period at different rates. access different current regimes or prey, to move away However, if one considers the development of critical from reefs at hatching, or to move between habitats at swimming speeds in terms of developmental age, settlement. High abundances of larval fishes have which provides a way of comparing the species at sim- been found in areas of internal waves, slicks and ilar stages of ontogenetic development (Job & Bell- fronts, and this may be due to active migration of lar- wood 2000), critical swimming ability increased simi- vae to these areas, possibly because of increased food larly for all 3 species. Interestingly, Job & Bellwood availability (Franks 1992). Critical swimming capabili- describe a comparable pattern in the development of ties of larval fishes are likely to be important for such visual abilities of these species. small-scale movement, and possibly in the modifica- The differences between the swimming abilities of tion of broad-scale dispersal patterns via vertical the 3 species can also be explained by the morphology migration (see Job & Bellwood 2000). of the larvae. All 3 species display a similar pattern, In contrast to critical swimming speed, sustained with critical swimming speeds increasing with increas- swimming increased very differently among the spe- ing relative propulsive area. Other studies have re- cies with both age and size, suggesting that the factors ported that various measures of body muscle or propul- involved are more complex than those required for the sive area may, along with length, explain differences development of critical swimming abilities. The devel- in swimming abilities. These include differences in the opment of sustained swimming ability varied from no percentage of muscle relative to other body tissue development at all in Sphaeramia nematoptera, a fairly (Bainbridge 1960) and the sagittal area of the body smooth transition in Amphiprion melanopus, to a rapid musculature (Ryland 1962). Furthermore, the length of jump in Pomacentrus amboinensis. The development the body that is undulated is known to vary among of sustained swimming may be dependent upon both species (Jayne & Lauder 1995) and may explain differ- physiological and behavioural changes during onto- ences in swimming ability. Propulsive area, as defined geny. Sustained swimming experiments provide a in the present study, approximates the area of the fish measure of the long-term swimming abilities of larvae. that can be undulated during swimming. This variable However, this method may underestimate the swim- appears to give a good indication of the swimming ca- ming capabilities of larvae in the field (Stobutzki & pabilities of larval fishes, probably because it provides Bellwood 1997). In the trials, larvae were not permitted a measure of the development of the functional region to feed or rest and were unable to select their optimal of the fish used for swimming. If the relationship be- swimming speeds. Leis & Carson-Ewart (1998) have tween relative propulsive area and swimming ability is observed larvae feeding whilst maintaining such high similar in a number of additional species, it is possible speeds. Given food, rest and the ability to select pre- that this variable may prove useful for general predic- ferred swimming speeds, the swimming times might tions of the swimming capabilities of larval fishes, in- have been much higher. Although there is a great cluding taxa which are difficult to rear in captivity. Fisher et al.: Development of swimming abilities in reef fish larvae 171

Reynolds numbers (Re) have also been identified as suggests that there is a high potential for this species to an important factor in the development of swimming remain near to the natal reef. Preliminary results for abilities in larval fishes (Webb & Weihs 1986). Re val- Amphiprion percula (Fisher unpubl. data) indicated ues give a measure of the ratio of forces arising from that the rapid development of swimming ability ob- inertial and viscous effects as a fish moves through served in A. melanopus may be common to other water. At small Re values (<200), the interaction anemonefish species. These findings are consistent between the fish and the water changes so that swim- with the observation that there appears to be a high ming becomes energetically expensive. Considering degree of genetic differentiation between some popu- that all the species in this study did not develop sus- lations of anemonefish, an indication of limited disper- tained swimming until they had attained Re values sal (Bell et al. 1982). Job & Bellwood (2000) suggest considerably larger than 200, this may be a constrain- that the precocious development in anemonefishes ing factor only in the development of sustained swim- may be related to low post-settlement mortality rates ming in very young larvae. It is unlikely that Re values and the need for well-developed sensory systems are important for the development of sustained swim- rather than size at settlement. ming ability, particularly as all 3 species developed In at least 2 of the 3 species investigated, swimming sustained swimming at very different values of Re. abilities develop at an early enough age to substan- tially influence their dispersal patterns. If they use their swimming capabilities to modify dispersal, Implications for dispersal genetic evidence suggests that they are using them to retard rather than enhance dispersal (Doherty et al. The 3 species investigated in this study displayed 1995). There is considerable evidence from hydrody- quite different patterns of development of swimming namic studies suggesting that young larvae will not ability. Although some species, such as Sphaeramia necessarily be transported very far from reefs (Black et nematoptera, appear to exhibit very little swimming al. 1991, Black 1993), particularly in the summer ability, and may be considered ‘passive’, others are months (Frith et al. 1986). Local retention, along with clearly capable of exerting considerable influence over the early development of swimming ability in some their dispersal patterns. The 2 species that did exhibit demersal-spawning reef fishes, suggests that many sustained swimming, however, developed these abili- populations of reef fishes may have a significant poten- ties considerably faster than expected. Critical swim- tial for self-seeding1. ming abilities for these species also developed rapidly. The 3 species investigated in this study were all Pomacentrus amboinensis rapidly develops excellent demersal brooders. Fish larvae which hatch from dem- sustained swimming abilities somewhere between ersal eggs are more developed in terms of their sensory Days 12 and 15 after hatching. Thus, for nearly half of and swimming abilities than larvae which hatch from the larval phase, this species is capable of significantly pelagic eggs (Kobayashi 1989, Leis 1991a), and are influencing its dispersal patterns by active swimming. more likely to be able to remain near the reef of origin This changes the potential distance that larvae of this (Leis 1991b, Brogan 1994). Apparently anomalous dis- species are able to cover before settlement from about tribution patterns have been observed in larvae with 34 km (based on the ability to swim for the last 2 d of demersal eggs, with studies often reporting higher the pelagic period) to 93 km (based on the ability to abundances near to shore (Leis 1986, Kingsford & swim for the last 6 d of the pelagic period, at a speed of Choat 1989). This may indicate that these larvae are 13.5 cm s–1). These values may be considerably greater using the early development of their motor skills to if we further assume that larvae are able to delay meta- influence their dispersal patterns or that they are very morphosis until they find a suitable settlement site (see adept at using currents to reduce advection. It is yet to McCormick 1999) and that they are able to feed whilst be determined how swimming ability develops in swimming (see Leis & Carson-Ewart 1998). Amphi- pelagic spawners, but it is likely to follow a period of prion melanopus also develops sustained swimming relative passivity during early development. abilities earlier than expected, but differs from P. Larval fishes exhibit excellent swimming abilities amboinensis in increasing its sustained swimming that develop, especially for critical swimming ability, ability gradually with age. A. melanopus was capable early in larval life and improve consistently throughout of some sustained swimming starting between 3 and 5 d after hatching, and this ability increased consis- 1 tently throughout the larval period. The development Since submission of this paper 3 articles have been published highlighting the potential for self seeding in reef fishes. of significant swimming abilities relatively soon after See Jones et al. (Nature 402:802–804, 1999), Swearer et al. hatching, along with well-developed sensory systems (Nature 402:799–802, 1999) and Cowen et al. (Science 287: (Job & Bellwood 2000), and a very short larval duration 857–860, 2000) 172 Mar Ecol Prog Ser 202: 163–173, 2000

the larval period. However, swimming is only one Brett JR (1964) The respiratory metabolism and swimming aspect of larval biology that will enable them to exhibit performance of young sockeye salmon. J Fish Res Board behaviour potentially important in influencing disper- Can 21:1183–1226 Brogan MW (1994) Distribution and retention of larval fishes sal. Job & Bellwood (2000) indicate that larvae of the near reefs in the Gulf of California. Mar Ecol Prog Ser 115: study species are able to feed at considerable depths 1–13 (50 to 200 m) even from an early age, and that this may Choat JH, Doherty PJ, Kerrigan BA, Leis JM (1993) A com- provide access to a range of current regimes. The comparison of towed nets, purse seines and light aggregation devices for sampling larvae and pelagic juveniles of coral bination of prolonged access to deep water and the reef fishes. Fish Bull US 91:195–209 early development of swimming abilities suggest that Cowen RK, Hare JA, Fahay MP (1993) Beyond hydrography – even the youngest larvae of these taxa have consider- can physical processes explain larval fish assemblages able potential to actively modify their dispersal and within the Middle Atlantic Bight. Bull Mar Sci 53:567–587 reduce advection from reefs. Clearly, the potential of Doherty PJ (1991) Spatial and temporal patterns in recruitment. In: Sale PF (ed) The ecology of fishes on coral reefs. larvae to modify their dispersal is not restricted to the Academic Press, San Diego, p 261–293 last few days of the pelagic period. Doherty PJ, Planes S, Mather P (1995) Gene flow and larval The present study has demonstrated that there are duration in seven species of fish from the Great Barrier great differences in the development of swimming Reef. Ecology 76:2373–2391 Drucker EG (1996) The use of gait transition speed in com- ability in the larvae of reef fishes and that these differ- parative studies of fish locomotion. Am Zool 36:555–566 ences are apparent shortly after hatching. In order to Franks PJS (1992) Sink or swim. Accumulation of biomass at fully understand the potential capabilities of larvae in fronts. Mar Ecol Prog Ser 82:1–12 modifying their dispersal, the whole of the larval Frith CA, Leis JM, Goldman B (1986) Currents in the Lizard period must be considered. Information based solely Island region of the Great Barrier Reef Lagoon and their relevance to potential movements of larvae. Coral Reefs 5: on the swimming abilities of older stages may be mis- 81–92 leading because it does not take into account how Jayne BC, Lauder GV (1995) Speed effects on midline kine- quickly or slowly their relative swimming abilities matics during steady undulatory swimming of large mouth develop during the pelagic period. Overall, it appears bass, Micropterus salmoides. J Exp Biol 198:585–602 Job S, Bellwood DR (1996) Visual acuity and feeding in larval that there is no such thing as a typical fish larva at any Premnas biaculeatus. J Fish Biol 48:952–963 stage and that some species of fish have considerably Job S, Bellwood DR (2000) Light sensitivity in larval fishes: more potential for influencing their dispersal than implications for vertical zonation in the pelagic zone. Lim- others. nol Oceanogr 45:362–371 Kingsford MJ, Choat JH (1989) Horizontal distribution patterns of presettlement reef fish: are they influenced by the proximity of reefs? Mar Biol 101:285–297 We thank I. Stobutzki for use of her Acknowledgements. Kobayashi DR (1989) Fine-scale distribution of larval fishes: swimming channel; the staff of the James Cook University patterns and processes adjacent to coral reefs in Kaneohe Research Aquarium and D. Ross for technical support; R. Bay, Hawaii. Mar Biol 100:285–293 Alford, M. Sheaves, M. Blows and R. Brooks for statistical Leis JM (1986) Vertical and horizontal distribution of fish lar- advice; and R. Alford and R. Rowe for use of technical equip- vae near coral reefs at Lizard Island, Great Barrier Reef. ment. This paper was improved as a result of the comments of Mar Biol 90:505–516 J. Leis and M. McCormick. This research was supported by Leis JM (1991a) The pelagic stage of reef fishes: the larval an augmentative research grant from the CRC Reef Research biology of coral reefs. In: Sale PF (ed) The ecology of fishes Centre (R.F.) and the Australia Research Council (D.R.B). on coral reefs. Academic Press, San Diego, p 183–229 Leis JM (1991b) Vertical distribution of fish larvae in the Great Barrier Reef Lagoon, Australia. Mar Biol 109: LITERATURE CITED 157–166 Leis JM, Carson-Ewart BM (1997) In situ swimming speeds of Bainbridge R (1960) Speed and stamina in 3 fish. J Exp Biol the late pelagic larvae of some Indo-Pacific coral-reef 370:129–153 fishes. Mar Ecol Prog Ser 159:165–174 Bell LJ, Moyer JT, Numachi K (1982) Morphological and Leis JM, Carson-Ewart BM (1998) Complex behaviour by genetic variation in Japanese populations of the anemone- coral-reef fish larvae in open-water and near-reef pelagic fish Amphiprion clarkii. Mar Biol 72:99–108 environments. Environ Biol Fish 53:259–266 Bellwood DR, Leis JM, Stobutzki IC (1998) Fishery and reef Leis JM, Stobutzki IC (1997) Swimming performance of management. Science 279:2021–2022 late pelagic larvae of coral-reef fishes: in situ and labo- Black KP (1993) The relative importance of local retention ratory-based measurements. Proc Indo-Pac Fish Conf 5: and inter-reef dispersal of neutrally buoyant material on 575–583 coral reefs. Coral Reefs 12:43–53 Leis JM, Sweatman HPA, Reader SE (1996) What the pelagic Black KP, Moran PJ, Hammond LS (1991) Numerical models stages of coral reef fishes are doing out in blue water: day- show coral reefs can be self-seeding. Mar Ecol Prog Ser time field observations of larval behavioural capabilities. 74:1–11 Mar Freshw Res 47:401–411 Booth DJ, Baretta GA (1994) Seasonal recruitment, habitat McCormick M (1999) Delayed metamorphosis of a tropical associations and survival of pomacentrid reef fish in the reef fish (Acanthurus triostegus): a field experiment. Mar US Virgin islands. Coral Reefs 13:81–89 Ecol Prog Ser 176:25–38 Fisher et al.: Development of swimming abilities in reef fish larvae 173

Roberts CM (1997) Connectivity and management of swimming abilities of pre- and post-settlement coral reef Caribbean coral reefs. Science 278:1454–1457 fishes. J Exp Mar Biol Ecol 175:275–286 Roberts CM (1998) Fishery and reef management. Science Stobutzki IC, Bellwood DR (1997) Sustained swimming abili- 279:2021–2022 ties of the late pelagic stages of coral reef fishes. Mar Ecol Ryland JS (1962) The swimming speeds of plaice larvae. J Exp Prog Ser 149:35–41 Biol 40:285–299 Webb PW, Weihs D (1986) Functional locomotor morphology Sale PF, Cowen RK (1998) Fishery and reef management. Sci- of early life history stages of fishes. Tran Am Fish Soc 115: ence 279:2021–2022 115–127 Silk J (1976) A comparison of regression lines using dummy Williams DMcB, Wolanski E, Andrews JC (1984) Transport variable analysis: models and examples of the bivariate mechanisms and their potential movement of planktonic case. University of Reading larvae in the central region of the Great Barrier Reef. Stobutzki IC (1998) Interspecific variation in sustained swim- Coral Reefs 3:229–236 ming ability of late pelagic stage reef fish from two fami- Wolanski E, Doherty P, Carleton J (1997) Directional swim- lies (Pomacentridae and Chaetodontidae). Coral Reefs 17: ming of fish larvae determines connectivity of fish popula- 111–119 tions on the Great Barrier Reef. Naturwissenchaften 84: Stobutzki IC, Bellwood DR (1994) An analysis of the critical 262–268

Editorial responsibility: John Austin (Assistant Editor), Submitted: November 2, 1999; Accepted: February 15, 2000 Oldendorf/Luhe, Germany Proofs received from author(s): August 3, 2000